Priority timeline for the PromptFlow™ System.

May 29, 2023 Provisional patent

Provisional patent application 63/504,767 filed.

This provisional established priority for the system architecture: the interpreter for managing execution of code/prompt processes, the knowledge graph of reusable processes, embedding-based retrieval for composition, runtime-statistics-based process selection, and the monetization framework for re-used processes.

This priority date is the anchor for all subsequent claims.

Feb 16, 2024 Continuation

Second provisional application 63/554,642 filed.

Extended and clarified the original claims, refining the specification of the knowledge graph structure, embedding-based composition mechanism, and settlement flows for multi-party executions.

May 2024 Non-provisional

Non-provisional US Application 18/677,796 filed.

Published as US20240403005A1. Currently under examination. Covers the full system architecture, including the interpreter's role in orchestrating execution, the typed composition interfaces, and the monetization graph.

View the published patent on Google Patents ↗

2025 Notice of Allowance

USPTO Notice of Allowance on the PromptFlow™ mark.

The United States Patent and Trademark Office issued a Notice of Allowance for the PromptFlow trademark, covering downloadable computer software featuring a programming language for prompting neural networks (IC 009) and the corresponding online non-downloadable software (IC 042). Use-in-commerce specimen filing follows.

View status on USPTO TSDR (case 97913988) ↗

Nov 2024 For context

Anthropic introduces the Model Context Protocol — 18 months after the priority date.

The MCP protocol — now the de facto AI integration standard adopted across OpenAI, Google DeepMind, and Microsoft — emerged substantially after the PromptFlow™ System's claims were filed. The broader agent-framework and tool-composition patterns the industry has since converged on also emerged in this window.

PromptFlow™ is not a response to these developments. It's an independent architecture documented before the ecosystem existed, and the trust/settlement/composition layer that the ecosystem still doesn't have.

Trademark

PromptFlow

PromptFlow™ is a trademark of PromptFlow Technologies LLC, covering the following goods and services per the USPTO Notice of Allowance:

IC 009 — Downloadable software

Downloadable computer software featuring a programming language for prompting neural networks; computer software files in the nature of downloadable electronic data files featuring programming data in the field of AI.

IC 042 — SaaS / online software

Providing online non-downloadable computer software featuring a programming language for prompting neural networks; computer software files in the nature of online non-downloadable electronic data files featuring programming data in the field of AI.

What makes the PromptFlow™ System wholly different.

The PromptFlow™ System is not another workflow tool, orchestration framework, or integration protocol. It is structurally and functionally distinct from anything else in the space. Five properties define what's different:

01

It does not compete with the tools it federates.

MCP servers, n8n workflows, LangChain agents, hosted models — these are the substrate PromptFlow™ sits on top of, not competitors to it. The PromptFlow™ System is the measurement, composition, and settlement layer above them. Other AI tools are peers participating in the graph.

02

Every execution produces evidence.

The graph stores measured outcomes from real executions, not claims about capabilities. Rankings are computed from evidence — completion rate, cost, latency, schema conformance, audit verdicts. Every capability accumulates a quantified track record that users and other capabilities can route against.

03

Compositions earn money for everyone involved.

When a process executes, every contributor whose work was used — the composer, sub-process authors, MCP operators, eval case authors, auditors — receives a share proportional to contribution. Most AI tools monetize the platform operator; the PromptFlow™ System monetizes the graph's contributors.

04

The graph simplifies itself as models improve.

When a new model can do in one call what previously required a multi-step composition, the older path becomes alternate rather than default. Complexity sheds automatically as underlying models strengthen. This simplification mechanism — described in the original patent claims and validated empirically — is unique to the PromptFlow™ System.

05

Open specification, protected system.

The .promptflow file format and the specification are CC-BY-4.0. Reference implementations are MIT-licensed. The trademark and the underlying system architecture are protected. This combination is deliberate: the specification is meant to federate openly; the protection is meant to prevent any single actor from capturing the graph.